Concurrency: How Many Chats Should One Agent Handle at Once?
Concurrency is the number of chats, messages or cases an agent works on at the same time. It's most common in live chat and digital messaging, where one agent can hold several conversations while each customer reads and types.
A phone agent handles one customer at a time, so voice has a concurrency of one.
Concurrency is often sold as the reason digital channels are cheaper: two chats at once must need half the staff. It doesn't work like that.
Each conversation slows down as the agent juggles more, so the saving is smaller than the maths suggests, and the customer pays for it in waiting time you won't see in a queue report.
Why it matters
Concurrency is one of the biggest inputs into how many people you need for chat and messaging.
Get it wrong and your staffing plan, your response times and your customer experience are all wrong with it.
Why it's tricky
Raising concurrency always looks good on a spreadsheet. The cost shows up later, in slower replies, more errors, lower satisfaction and tired agents.
The right level depends on the contact, the channel and the person.
What this guide covers
A clear definition, the four numbers that make up concurrency, and what it really does to staffing.
Then how to set it, how to measure whether it's working, and the mistakes to avoid.
What is Concurrency?
In a contact centre, concurrency describes how many interactions an agent has open and is actively working at one time. An agent with three live chats open has a concurrency of three at that moment.
Over a shift, the figure that matters for planning is the average: how many conversations the agent was really working on at once, across all the time they were handling contacts.
✓ Concurrency is
- How many chats, messages or cases one agent works at the same time
- Usually set as a maximum (the cap) in your chat or messaging platform
- A key input into staffing for digital channels
- Something that changes the customer's experience, not just the cost
✕ Concurrency is not
- A straight multiplier on agent capacity
- The same as Occupancy, which measures how busy an agent is
- One number that suits every contact type and every agent
- The number of simultaneous calls your phone system can carry (see below)
Plain-English Definition
Concurrency is how many customers an agent is talking to at the same time.
On the phone it's always one. In live chat it's often two or three. In messaging, where customers reply when it suits them, it can be much higher.
The higher it goes, the longer each customer waits for the next reply.
Not to be confused with concurrent calls or licences
Telephony and software providers also use "concurrency" to mean how many calls a phone line, SIP trunk or platform can carry at once, or how many users can be logged in at the same time.
That's a capacity and licensing question for your systems. This guide is about agent concurrency: how many conversations one person works at once.
Why Concurrency Matters in CX
Concurrency sits at the point where cost, quality and customer experience pull against each other. That's why it deserves a deliberate decision, not a default setting.
For CX leaders
Customers experience concurrency as gaps — waiting for the next reply in a conversation that was meant to be live.
Too high, and live chat starts to feel like slow email. That gap is where satisfaction drops and customers give up mid-conversation.
For contact centre leaders
Concurrency is one of the main reasons digital contacts cost less than calls, and one of the main ways the saving gets overstated.
It also changes what your Average Handling Time and Occupancy figures mean for digital channels, so they can't be read the same way as voice.
For workforce planners
Staffing models built for voice assume one contact at a time. Chat with concurrency breaks that assumption.
Planning digital channels needs a concurrency-aware approach — the WFM Hub brings together the planning tools and guidance to do it.
What the research says
Sources: Comm100 Live Chat Benchmark Report 2025 (analysis of 220+ million chats); [24]7 Customer, "Chat and the Concurrency Myth" (2013).
Reading these numbers honestly
Treat these as context, not a target. Both sources are chat software or service providers, and the [24]7 research is more than a decade old.
The pattern they share is the useful part. [24]7 found customer satisfaction and first contact resolution both rose as concurrency fell, and Comm100 warns that raising concurrency without considering experience and complexity leads to rushed conversations and lower satisfaction.
The only number that counts is the one your own response times and quality results support.
The Four Numbers Behind Concurrency
"Our concurrency is three" can mean very different things. Separate these four numbers and the conversation gets much clearer.
The cap
The maximum number of conversations the platform will route to an agent at once. It's a setting, not a result.
Many platforms let you set different caps by agent, skill or contact type.
Average concurrency
How many conversations agents actually worked at once, on average, across their handling time.
It's almost always below the cap, because demand rises and falls. This is the number to plan with.
Conversation duration
How long each chat stays open from start to finish. It stretches as concurrency rises, because the agent's attention is shared.
A chat that takes eight minutes on its own may take twelve when the agent is juggling two.
Response time in the conversation
How long the customer waits for each reply once the conversation has started.
Queue reports rarely show it, yet it's what the customer feels most. It's the first thing to suffer when concurrency is too high.
Concurrency by channel
| Channel | Typical concurrency | Why |
|---|---|---|
| Voice | One | Real-time conversation that needs full attention |
| Live chat | Often two to three | Synchronous Communication — the customer is waiting on the other end for each reply |
| Messaging (WhatsApp, SMS, in-app) | Can be much higher | Asynchronous Communication — the customer replies when it suits them, so conversations pause |
| Email and cases | Usually one at a time | Worked in sequence from a backlog, rather than held open together |
The point most staffing models miss
Concurrency only saves capacity if the extra conversations don't stretch the ones already open.
They always do, to some degree. So doubling concurrency never halves the staff you need — and past a certain point it saves nothing at all while making every conversation slower.
How to Set the Right Concurrency
There is no universal right number. The right cap is the one your response times, quality results and customer feedback support for each type of contact.
Start with the type of contact
Simple, predictable requests such as order status or opening hours can tolerate a higher cap. Complaints, hardship, technical faults and sales conversations need more attention per customer.
Set caps by queue or contact type, not one cap for the whole team.
Match it to the agent
New agents should start at one or two conversations and step up as their accuracy and response times hold.
Some experienced agents handle more well; some never do. Juggling written conversations is a skill that needs training — Live Chat Customer Service covers it at agent level.
Treat live chat and messaging differently
A live chat customer is sitting there waiting. A messaging customer may reply in an hour.
Applying a messaging-style cap to live chat quietly turns a live channel into a slow one.
Step up slowly and watch the customer measures
Change the cap one step at a time and compare response time within the conversation, chat duration, satisfaction and resolution before and after.
If response gaps lengthen or satisfaction drops, step back.
Plan staff on stretched durations
Measure how long conversations last at each concurrency level and plan with those figures, divided by average concurrency — never the cap.
Standard Erlang C models assume one contact at a time, so they need adjusting for chat. Erlang C isn't broken, but using it for the wrong job is.
Use tools that lighten the load
Good knowledge articles, saved responses that agents personalise and AI assistance can make higher concurrency workable by reducing the effort per reply.
Add the tools first, then test whether a higher cap holds. Raising the cap and hoping doesn't work.
Practical example: what doubling concurrency really saves
The figures below are illustrative. They show how stretched conversations change the maths; your own durations will differ.
| Average concurrency | Chat duration | Agent hours for 600 chats | Saving vs one at a time |
|---|---|---|---|
| 1 | 8 minutes | 600 × 8 ÷ 1 ÷ 60 = 80 | — |
| 2 | 12 minutes | 600 × 12 ÷ 2 ÷ 60 = 60 | 25%, not 50% |
| 3 | 18 minutes | 600 × 18 ÷ 3 ÷ 60 = 60 | No further saving, and slower replies |
What the example shows
Moving from one to two conversations saved a quarter of the hours, which is real and worth having.
Moving from two to three saved nothing, because each chat stretched by as much as the extra slot added. The customers simply waited longer between replies.
Benefits of Getting Concurrency Right
Staffing plans that hold up
Planning on real average concurrency and real durations gives you numbers that match the day.
No more rosters built on a cap the team never reaches.
Live chat that feels live
Customers get replies quickly enough that the conversation flows.
That's the whole reason they chose chat over email.
Better quality and resolution
Agents with enough attention per customer make fewer mistakes and resolve more in one conversation.
That means fewer repeat contacts later.
Honest cost comparisons
You can compare the real cost of chat, messaging and voice, rather than an inflated digital saving.
Channel decisions get better as a result.
Less agent burnout
Constant task switching across too many conversations is exhausting.
Caps matched to the person and the work keep the job sustainable.
Smarter routing
Different caps by contact type let simple requests flow at higher concurrency.
Complex and sensitive conversations get the focus they need.
Common Pitfalls with Concurrency
Raising the cap to hit a cost target
Increasing concurrency is the quickest way to make a digital channel look cheaper. It's also the quickest way to make it slower and less accurate.
The fix: only raise a cap when response time, quality and satisfaction hold at the higher level.
Planning with the cap instead of the average
A cap of three doesn't mean agents work three conversations all day. Staffing on the cap leaves you short every time demand dips between conversations.
The fix: plan with measured average concurrency and the durations that go with it.
Ignoring the wait inside the conversation
Your queue report may show customers connected to an agent within seconds. It won't show them waiting two minutes for every reply once they're connected.
The fix: report agent response time within the conversation alongside queue wait time.
One cap for every contact and every agent
A single setting treats a password reset and a complaint as the same workload, and a first-week agent the same as a five-year veteran.
The fix: set caps by contact type and step agents up as their results allow.
Running digital staffing through a voice model
Classic Erlang C assumes one contact per agent at a time. Feeding it chat volumes without adjusting for concurrency produces a staffing number that's wrong in both directions. Understand what Erlang C assumes before you use it.
The fix: use a concurrency-aware capacity model for digital channels and validate it against what actually happens.
Reading digital AHT like voice AHT
A chat's handle time includes all the time the agent was working other chats. Comparing it with call handle time, or coaching on it alone, misleads.
The fix: always report digital handle time together with the concurrency it was achieved at.
Warning: If you raised concurrency and your cost per chat fell while satisfaction, resolution or customers finishing their chats also fell, the saving isn't real.
You've moved the cost onto customers and into repeat contacts.
How to Measure Concurrency — and Know It's Working
Concurrency isn't a number to push as high as possible. Measure what it does to customers and agents.
Track average concurrency against the cap
Report both for each queue and team. A large gap means demand or routing isn't filling the slots.
Average concurrency sitting at the cap all day means agents may be overloaded.
Measure response time within the conversation
Track the first agent response and the average gap between customer messages and agent replies.
This is the clearest early warning that concurrency is too high.
Compare outcomes by concurrency level
Split CSAT, resolution and quality scores by how many conversations the agent had open at the time.
If results fall away at a particular level, that's your practical ceiling for that contact type.
Watch customers who leave mid-conversation
Customers who stop replying or close the chat before it's resolved are often reacting to slow replies.
A rise after a concurrency change tells you something.
Ask the agents
Agents know when they're juggling too much long before the reports do.
Regular feedback on workload, alongside error rates, protects quality and retention.
The rule
Set concurrency by what customers experience, not by what the cost model wants.
The right level is the highest one where response time, quality and resolution hold — and it will be different for different contacts and different agents.
How fast do your chats and emails really get answered?
ACXPA's Customer Service Benchmarking measures what a real customer gets when they email or open a chat with you, scored independently against the Contact Centre CX Standards.
Frequently Asked Questions
What is concurrency in a contact centre?
Concurrency is the number of chats, messages or cases an agent works on at the same time.
It applies mainly to live chat and digital messaging. Voice calls have a concurrency of one.
How many chats can an agent handle at once?
Many live chat teams work at two or three, with newer agents at one or two. Messaging, where customers reply in their own time, can support more.
The right number depends on the complexity of the contact and the agent's experience. Test it against your own response times and quality results.
Should we set a concurrency target?
No. Set a cap for each contact type, not a target to push towards.
A target encourages agents and planners to fill every slot regardless of what it does to customers. Judge the cap by response time, quality and resolution.
Does higher concurrency mean you need fewer staff?
Up to a point. Each conversation stretches as concurrency rises, so the saving is always smaller than the multiplier suggests.
Past a certain level there is no saving at all, only slower replies.
Is concurrency the same as occupancy?
No. Occupancy is the share of an agent's available time spent handling contacts. Concurrency is how many contacts they handle at once.
Concurrency makes occupancy harder to read for digital channels, because an agent with one chat open in a three-chat cap still counts as occupied.
Can I use an Erlang C calculator for chat?
Not as it stands. Erlang C assumes each agent handles one contact at a time, which suits voice.
For chat you need to adjust for concurrency and stretched durations. The Erlang C Calculator remains the right starting point for voice staffing.
What's the difference between concurrency in live chat and messaging?
Live chat is synchronous: the customer is waiting for each reply, so concurrency has to stay low enough to keep replies quick.
Messaging is asynchronous: conversations pause while customers do other things, so agents can hold more open without the customer noticing.
What does concurrency mean for phone systems?
In telephony and software licensing, concurrency means how many calls a line or platform can carry at once, or how many users can be logged in together.
It's a separate meaning from agent concurrency, which is about how many conversations one person handles.
Where to Next
Summary
Concurrency is the number of chats, messages or cases an agent works on at the same time. Voice is always one; live chat is often two or three; messaging can go higher because customers reply in their own time.
It's one of the biggest inputs into digital staffing, and one of the most often overstated.
Every extra conversation stretches the ones already open. Doubling concurrency never halves the staff you need, and beyond a certain point it saves nothing while making every reply slower.
That cost lands on customers as waiting inside the conversation, which queue reports don't show.
Set caps by contact type and agent, plan on measured average concurrency and stretched durations, and judge the result by response time, quality and resolution.
The right concurrency is the highest level your customers don't notice.
